N503 Non-Experimental Wk 5

Dr. K KovacsBurns 1

Non-experimental Research Designs

NURSING 503 Week 5, October 5, 2005

Dr. K KovacsBurns 2

Overview – Main topics

Non-experimental vs experimental designs

When to use non-experimental designs The types of non-experimental designs Advantages & disadvantages Studies using surveys Epidemiologic research designs

Dr. K KovacsBurns 3

A continuum of quantitative research design

l_____________l_____________lNon-experimentalQuasi-experimentalExperimental

Dr. K KovacsBurns 4

Reasons for Undertaking Nonexperimental Studies

Number of human characteristics/ independent variables are not subject to experimental manipulation or randomization

Some variables cannot ethically be manipulated For some research, it is not practical to conduct a

true experiment/manipulate variables For some situations, it is more realistic to explore

phenomena in more natural manner Nonexperimental research is often needed to scope

out the experimental one

Dr. K KovacsBurns 5

Key Characteristics

A research design in which a researcher observes a phenomenon without manipulating the independent variables(s)

No manipulation Independent variables have already

occurred, so no control over them Clear, concise problem statement that is

based on a theoretical framework, or natural phenomenon.

Dr. K KovacsBurns 6

Classification of

NonExperimental designs:

Descriptive/exploratory survey studies Interrelationship/difference

Correlational studies Ex post facto studies Prediction studies Developmental studies

Cross-sectional & longitudinal studies Retrospective & prospective studies

Dr. K KovacsBurns 7

Descriptive/Exploratory Survey Broadest category Detailed observations, descriptions &

documentation of existing variables Little is known about the phenomenon Justifies, assesses current conditions/practice Variables of interest: opinions, attitudes or facts Determines differences between variables Data collected by questionnaire or interview Researchers only relate one variable to another,

no attempt to determine causation

Dr. K KovacsBurns 8

Exploratory Survey Research

Designed to obtain information about prevalence distribution interrelations of variables within a population

Census vs. sample surveys Self-reporting Flexibility and broadness Superficiality – extensive vs. intensive

analysis

Dr. K KovacsBurns 9

Two aspects about surveys:

Survey Content Direct questioning Answering how,

what,and to what extent questions

Usually focus on what was done and what people plan to do in the future

Survey Administration Different data

collection methods Personal interviews Telephone interviews Self-administered

questionnaires (SAQs)

Mixed-mode strategy

Dr. K KovacsBurns 10

How surveys inform practice Needs assessment – estimate needs of a group,

community or organization; & make recommendations for priorities

Face-to-face interviews using surveys yields quality data about the person being interviewed & the question posed

Delphi surveys for short-term forecasting based on panel of experts’ opinions, predictions, and judgment about a topic

Secondary analysis of surveys yields more details Meta-analysis of similar studies to integrate large

body of findings & observations


Descriptive Research

Purpose is to observe, describe, & document aspects of a

situation as it naturally occurs serve as a starting point for hypothesis

generation or theory development Types:

Descriptive Correlational Studies Univariate Descriptive Studies

Prevalence Studies Incidence Studies


Descriptive Correlational Studies

Describes the relationship among variables rather than infer cause-and-effect relationships

Are usually cross-sectional


Univariate Descriptive Studies

Could focus on one or more variables Undertaken to describe the frequency

of occurrence of a behavior or condition or each variable rather than relationships between or among them

Types: Prevalence Studies Incidence Studies


Prevalence Studies

Done to determine the prevalence rate of some condition at a specific point in time

Data is obtained from the population at risk for the condition – cross sectional designPrevalence Rate (PR) = # cases with condition X K

# in population at risk


Incidence Studies

Used to measure the frequency of developing new cases

Need longitudinal designs

Incidence Rate (IR)= # new cases with condition over given period X K# at risk of becoming a new case

Relative Risk: an estimate of risk of “caseness” in one group vs. another; contribution of risk factors E.g. males vs females for acquiring depression


Descriptive/Exploratory Survey (cont’d)

Advantage: large amount of information can be obtained from a large population in aneconomical manner which is “surprisingly” accurate

Disadvantages Info tends to be superficial as breadth is

emphasized Expertise in: sampling techniques, questionnaire

construction, interviewing and data analysis to produce a reliable and valid study.

Time-consuming & sometimes costly


Interrelationship/Difference Studies: Correlational Studies

Examines if variables covary Quantifies the strength or relationship between the variables (not cause &

effect) +ve or –ve direction relationship determined Advantages:

Increased flexibility when investigating complex relationships among variables Efficient and effective method of collecting a large amount of data Potential for practical application in clinical settings Potential foundation for future, experimental studies Framework for exploring relationships that are not manipulated.

Disadvantages: Unable to manipulate variable of interest No randomization in sampling Generalizability decreased as dealing with preexisting groups Unable to determine a causal relationship because of the lack of manipulation,

control and randomization.


Interrelationship/Difference Studies: Ex Post Facto Studies

Literally means ‘from after the fact’ Also known as causal-comparative studies or

comparative studies Explores differences/relationships between variables

(similar to quasi-experimental designs) Advantages:

Allows for establishment of a differential effect Similar to correlational designs Offers a higher level of control

Disadvantages: Lack of control on variables Unable to draw causal linkage Problem of alternative hypothesis


Interrelationship/Difference Studies: Developmental Studies

Not only concerned with existing status & interrelationship of phenomena but also with changes from elapsed time.

Cross sectional (one/more time points, perhaps different groups) vs Longitudinal (several time points with same group over extended period)

Retrospective (dependent variable has already been affected by independent variable, link present events to past events) vs Prospective (link present events to presumed future effect, less common, considered stronger design)


Others

Natural Experiments: study of a group exposed to natural or other phenomenon that have health or other consequences, compared with a nonexposed group; people are affected at random

Path Analytic Studies: using a technique called path analysis, nonexperimental data is tested against a hypothesized causal inference


Epidemiologic Research Designs

Cohort studies: trend study in which specific subpopulations (eg age specific) are examined over time for generational differences prospective & retrospective

Case-control study: comparison of cases/subjects (with specific condition), with controls (without condition); only difference should be exposure to presumed cause

Cross-sectional design: phenomena under study are captured during one period of data collection; one point in time


Continuum of designs for inferring causality

StrongestWeakest

Design Non Experimental Designs Design

_______________________________________

True Experim

ental Design

Quasi E

xperimental D

esignP

re Experim

ental Design

Natural E

xperiment

Path A

nalytic Design

Prospective C

orrelationalR

etrospective Correlational

Descriptive D

esigns


Questions/Discussion

N503 Non-Experimental Wk 5

Documents

link present

correlational

descriptiveexploratory

interrelationshipdifference

large amount

experimental

time points

variables